Can the AI recommend products to customers?

Yes. In a live conversation the assistant can suggest products from your catalog, working from three signals: what the shopper says they want, what they've been browsing, and the page they have open right now.

How it picks

When someone describes what they're after, the assistant searches your real Shopify catalog and weighs stated preferences, like size, flavor, budget, skin type, or compatibility, alongside the product or collection page in their browser.

For open-ended requests ("what should I try?") and comparisons, it usually narrows to two or three strong options rather than listing every SKU you carry. A short list is something a person can decide from; a wall of links isn't. When the customer names a product type or a specific product by title, selection follows that intent instead.

What the shopper sees in chat

Every recommendation links to the product page on your store, so variants, price, and availability sit one click away and the path to purchase stays on your domain. In comparisons, the assistant calls out the differences that matter for the decision: format, ingredients, price tier, who each option suits. When the conversation has narrowed to one clear product, it can also show a single call-to-action button that opens that product page. Several products still in play stay as inline links; the assistant does not pick one of them as a lone button.

The guardrails

Recommendations come only from products that exist in your store. No invented SKUs, no brands you don't carry, no suggesting items that are unavailable. When catalog data doesn't support a confident match, the assistant asks a clarifying question, offers alternatives from what you do sell, or connects the shopper with your team. On high-trust categories, a wrong suggestion costs more than a slow reply, so it doesn't guess.

When a person should take over

Bespoke bundles, medical-adjacent claims you'd rather not automate, VIP buyers who expect a personal touch. The assistant can gather preferences and product context first, then escalate, so your staff starts with a clear brief instead of redoing the whole discovery conversation.

Browsing context does real work here. A shopper on a starter kit page who asks "which refill should I get?" gets refills that relate to that kit, not bestsellers from another category.